updated model details

README.md

@@ -1 +1,30 @@
-This repository contains models trained to predict orientation {0:'up', 1:'down'} of images. The dataset used is the CIFAR10, normalized using mean and stdev of the training set, and resized to 224x224.
+This repository contains model trained to predict orientation {0:'up', 1:'down'} of images.
+The model '.pkl' file contains a dictionary with the following keys:
+
+---
+'optimizer': optimizer state dictionary
+'scheduler': scheduler state dictionary
+'model': model state dictionary
+'epoch': checkpoint epoch
+---
+
+---
+framework: pytorch
+model: EfficientNet-B1
+dataset: CIFAR10 (restructured to have random upright and upside-down samples, with labels {0:'up', 1:'down'}
+imgSize: 224x224x3
+---
+
+The model can be initialized as:
+
+'''
+# download pretrained model
+model = torchvision.models.efficientnet_b1(pretrained=True)
+
+# define input and output features size
+in_features = model.classifier[1].in_features
+out_features = 2
+
+# replace the last layer
+model.classifier[1] = nn.Linear(in_features, out_features, bias=True)
+'''